Statistical heterogeneity can increase the uncertainty of results and reduce the quality of evidence derived from systematic reviews. At present, it is uncertain what the major factors are that account for heterogeneity in meta-analyses of analgesic adjuncts. Therefore, the aim of this review was to identify whether various covariates could explain statistical heterogeneity and use this to improve accuracy when reporting the efficacy of analgesics.METHODS:
We searched for reviews using MEDLINE, EMBASE, CINAHL, AMED, and the Cochrane Database of Systematic Reviews. First, we identified the existence of considerable statistical heterogeneity (I2 > 75%). Second, we conducted meta-regression analysis for the outcome of 24-hour morphine consumption using baseline risk (control group morphine consumption) and other clinical and methodological covariates. Finally, we constructed a league table of adjuvant analgesics using a novel method of reporting effect estimates assuming a fixed consumption of 50 mg postoperative morphine.RESULTS:
We included 344 randomized controlled trials with 28,130 participants. Ninety-one percent of analyses showed considerable statistical heterogeneity. Baseline risk was a significant cause of between-study heterogeneity for acetaminophen, nonsteroidal anti-inflammatory drugs and cyclooxygenase-2 inhibitors, tramadol, ketamine, α2-agonists, gabapentin, pregabalin, lidocaine, magnesium, and dexamethasone (R2 = 21%–100%; P < .05). There was some evidence that the methodological limitations of the trials explained some of the residual heterogeneity. Type of surgery was not independently associated with analgesic efficacy. Assuming a fixed baseline risk of 50 mg (in order of efficacy), gabapentin, acetaminophen, α2-agonists, nonsteroidal anti-inflammatory drugs and cyclooxygenase-2 inhibitors, pregabalin, tramadol, magnesium, and lidocaine demonstrated moderate clinically significant reductions (>10 mg). We could not exclude a moderate clinically significant effect with ketamine. Dexamethasone demonstrated a small clinical benefit (>5 mg).CONCLUSIONS:
We empirically identified baseline morphine consumption as the major source of heterogeneity in meta-analyses of adjuvant analgesics across all surgical interventions. Controlling for baseline morphine consumption, clinicians can use audit data to estimate the morphine-reducing effect of adding any adjuvant for their local population, regardless which surgery they undergo. Moreover, we have utilized these findings to present a novel method of reporting and an amended method of graphically displaying effect estimates, which both reduces confounding from variable baseline risk in included trials and is able to adjust for other clinical and methodological confounding variables. We recommend use of these methods in clinical practice and future reviews of analgesics for postoperative pain.